About

I am a Ph.D. student in Electrical Engineering at the University of Rhode Island, where I am advised by Dr. Yu Bi. My research sits at the intersection of machine learning and security: I design and evaluate backdoor attacks and defenses across convolutional networks, spiking neural networks (SNNs), large language models, and vision-language models, including retrieval-augmented (VLRAG) pipelines at million-image scale.

Before URI, I completed an M.S. at Saga University, Japan, working on ultrasonic Lamb-wave simulation for non-destructive testing, and a B.Sc. in Electrical and Electronic Engineering at KUET, Bangladesh.

• Adversarial ML & model security
• Backdoor attacks & defenses
• Spiking neural networks
• Vision-language models & RAG
• Computer vision

News

• 2026Lite-BD accepted at the IEEE/INNS International Joint Conference on Neural Networks (IJCNN 2026).
• 2026DeepCurer published in Neurocomputing.
• 2026NoiseAttack accepted at the International Journal of Information Technology.
• 2026Serving as a reviewer for IJCNN 2026.
• 2024Started as a Graduate Teaching Assistant for Digital Circuit Design (ELE 202) and Digital Circuit Design with FPGAs (ELE 302).
• 2023Joined the University of Rhode Island as a Ph.D. student.

Selected Publications

1. Multi-Knowledge Poisoning Attack with Backdoored Retriever on RAG-based Vision-Language Models

   Abdullah Arafat Miah, Yu Bi

   In preparation, targeting NAACL 2027 (ARR Oct. 2026 cycle) In preparation

2. DeepCurer: Pruning-based Backdoor Mitigation via Progressive Neuron Ranking using Adversarial Proxies

   Abdullah Arafat Miah, Yu Bi

   Neurocomputing, 2026, art. 134409 Published

   DOI Code

3. Lite-BD: A Lightweight Black-box Backdoor Defense via Reviving Multi-Stage Image Transformations

   Abdullah Arafat Miah, Yu Bi

   IEEE/INNS International Joint Conference on Neural Networks (IJCNN), 2026 Accepted

   arXiv

4. T-Backdoor: Exploiting Temporal Redundancy in Neuromorphic Data for Spike-preserving Backdoor Attacks on SNNs

   Abdullah Arafat Miah, Kevin Vu, Yu Bi

   IEEE Transactions on Artificial Intelligence Under review

5. BadSNN: Backdoor Attacks on Spiking Neural Networks via Adversarial Spiking Neuron

   Abdullah Arafat Miah, Yu Bi

   Neurocomputing Under review

   arXiv

6. BadImplant: Injection-based Multi-Targeted Graph Backdoor Attack

   Md Nabi Newaz Khan, Abdullah Arafat Miah, Yu Bi

   Journal of Information Security and Applications Under review

   arXiv

7. NoiseAttack: An Evasive Sample-Specific Multi-Targeted Backdoor Attack Through White Gaussian Noise

   Abdullah Arafat Miah, Kaan Icer, Resit Sendag, Yu Bi

   International Journal of Information Technology Accepted

   arXiv

8. Exploiting the Vulnerability of Large Language Models via Defense-Aware Architectural Backdoor

   Abdullah Arafat Miah, Yu Bi

   SN Computer Science Under review

   arXiv

Earlier publications (2019 to 2021) on drowsiness detection, video-based sound extraction, and ultrasonic defect detection are listed on Google Scholar.

Selected Projects

• Perplexity-Based Defense Against LLM Backdoor Attacks

  Implemented a word-trigger LLM backdoor and a perplexity-based defense, showing defense efficacy degrades sharply as trigger length grows.

• GI-Tract Finding Classification with CNNs and ML Classifiers

  Classified endoscopic findings in the gastrointestinal tract by combining CNN feature extraction with classical machine-learning classifiers.

• Sentiment & Emotion Recognition from Text

  Built an attention-augmented LSTM for joint sentiment and emotion classification from text.

• Real-Time Drowsiness Detection using CNN + Eye Landmarks

  Fused eye-landmark distance features with a custom CNN for real-time video-based drowsiness detection; published in Springer AISC.

Teaching & Service

• Graduate Teaching Assistant, Dept. of Electrical, Computer & Biomedical Engineering, URI (Sep 2024 to present), lead undergraduate lab sections for Digital Circuit Design (ELE 202) and Digital Circuit Design with FPGAs (ELE 302), including lab preparation, instruction, grading, and exam proctoring.
• Reviewer, IEEE/INNS International Joint Conference on Neural Networks (IJCNN), 2026.

Education

• Ph.D., Electrical Engineering, AI/ML Security

  University of Rhode Island, Kingston, RI · 2023 to 2028 (expected) · GPA 3.95/4.00

• M.S., Electrical Engineering

  Saga University, Japan, Lamb-wave ultrasonic simulation for non-destructive testing

• B.Sc., Electrical & Electronic Engineering

  Khulna University of Engineering & Technology (KUET), Bangladesh